hierarchical clustering
a multistage procedure in which entities are grouped into ever larger and more heterogeneous clusters or separated into ever smaller and more homogeneous clusters (see cluster analysis). The most common form of hierarchical clustering is agglomerative clustering, in which the individual entities are first paired, then these pairs are paired, and so on until ultimately all entities form a single large group or cluster. A second form of hierarchical clustering is divisive clustering, in which all entities initially are part of one large group which is split into progressively smaller groups according to dissimilarity until ultimately each entity is in its own group or cluster. The outcome of both types of hierarchical clustering may be represented graphically as a dendrogram.